On Sun, 1 Jun 2008, Peter Kümmel wrote:
Kevin Atkinson wrote:
On Sun, 1 Jun 2008, Peter Kümmel wrote:
Kevin Atkinson wrote:
On Sun, 1 Jun 2008, Peter Kümmel wrote:
Kevin Atkinson wrote:
If so stop. Most asserts perform important safety checks and disabling
them could lead to data corruption and other nasty things. Also, they
should have a negligible effect on performance.
I'm not sure if this is correct, STL checks could be expensive even when
optimizing the code.
Can you please show me proof of this.
It is just a impression because I've never heard about such a strategy to
ensure correctness. I only know, in debug mode msvc-stl-code is several
times slower than in release mode, due to disabling optimizations and
executing additional checks.
But a proof would only be a aspell benchmark which I don't have.
I've tried to benchmark with lyx and msvc, but it is not possible with msvc
(which means nothing in a gnu project). There a different runtime libraries
for debug und release which must not be mixed, but using _DEBUG enforces
linking against both runtime libraries.
But I assume gcc doesn't have such limitations.
So are you saying in order not to link with the release runtime libraries
NDEBUG can not be defined. This in my option is extremely stupid.
If this is the case, I don't think I can ever officially support Aspell
compiled with msvc. At very least I won't be able to accept bug reports
for Windows binary built in release mode. I will expand on the reasons
latter if necessary.
_______________________________________________
Aspell-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/aspell-devel
